We have been working on the same problem with NT clients... And all we
could get out of IBM was
to reduce the TCP window size from 512 to 64K.. They couldnt come up with
an anserwer to why the 512K
was a problem..
"Richard C. Dempsey" <dempsey AT KODAK DOT COM> on 03/26/99 10:25:13 AM
Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
cc: (bcc: Jim Hunt/IS/Vencor)
Subject: Re: TCP Errors
Server is Solaris 2.6, ADSM 3.1.2.13, TCPWindowsize 512K
Client is Solaris 2.6, ADSM 3.1.0.6, TCPWindowsize is the default (32K)
I have since chatted with our network switch guru. The client HME
ethernet interface card was forced to 100 Mb/s full duplex, but the
switch had autonegotiated its interface to 100/half. We also observed
on the counters many, many collisions and multi-collisions (which
should not be happening on a switch). We have forced the switch i/f
to match the host at 100/full. In other cases, I have observed that
this dramatically improved network (and backup) throughput.
OTOH, I suppose it would not be a bad idea to have the client and
the server use the same values for TCPWindowsize, eh?
Rich
At 09:22 AM 3/26/99 -0500, you wrote:
>What is your TCP window size set to on the server and the clients? What
>clients are you tring to backup? NT AIX ???
>
>
>
>
>"Richard C. Dempsey" <dempsey AT KODAK DOT COM> on 03/26/99 08:45:08 AM
>
>Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
>
>To: ADSM-L AT VM.MARIST DOT EDU
>cc: (bcc: Jim Hunt/IS/Vencor)
>Subject: Re: TCP Errors
>
>
>
>
>Increasing COMMTimeout from 300 seconds to 2000 seconds allowed the backup
>to complete last night. However, about 5 sessions were lost and had to be
>re-established to pull it off. I believe that there is a network problem
>here, and I will be taking it up with our network switching guru when he
>gets in.
>
>Thanks for your help,
>Rich
>
>>Date: Thu, 25 Mar 1999 11:26:40 -0500
>>To: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
>>From: "Richard C. Dempsey" <dempsey AT kodak DOT com>
>>Subject: Re: TCP Errors
>>
>>PS: Out of band conversation with Richard Sims has pointed to the server
>>error message "ANR0481W Session 30 for node ALTS1 (SUN SOLARIS)
terminated
>>- client did not respond within 300 seconds." as indicating that
>COMMTimeout
>>is set way too low. We will test COMMT 2000 tonight.
>>
>>>Recently, a BigIP redirector was interposed between our ADSM server
>>>(3.1.2.13, Solaris 2.6) and a client (3.1.0.6, Solaris 2.6). Since
>>>then, the client has not had a successful scheduled incremental backup.
>>>The last line in the SchedLog is typically something like:
>>>
>>>03/24/99 20:58:50 Incremental backup of volume '/export/data3'
>>>
>>>where '/export/data3' will vary, but, of course, is always a mount
>>>point (filespace). Both dsmc and dsmadmc seem to run fine interactively
>>>from this client, but I have not tried an interactive incremental.
>>>Router ACLs prevent me from running the GUI. (I'm just a CLI kind of
>>>guy :). I interpret this as showing that communcations on port 1500
>>>between the server and the client are unimpeded by BigIP.
>>>
>>>I have included the end of ErrorLog below. In /usr/include/sys/errno.h,
>>>we learn that error 32 is 'Broken Pipe'. Does anybody have any ideas
>>>what could be causing this?
>>>
>>>Thanks,
>>>Rich
>>>
>>>03/24/99 15:31:21 TcpRead(): recv(): errno = 131
>>>03/24/99 15:31:21 sessRecvVerb: Error -50 from call to
>'readRtn'.
>>>03/24/99 20:00:42 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:00:42 sessSendVerb: Error sending Verb, rc: -50
>>>03/24/99 20:00:42 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:00:42 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:00:42 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:00:57 ANS1810E ADSM session has been reestablished.
>>>03/24/99 20:07:58 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:07:58 sessSendVerb: Error sending Verb, rc: -50
>>>03/24/99 20:07:58 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:07:58 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:07:58 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:08:13 ANS1810E ADSM session has been reestablished.
>>>03/24/99 20:41:49 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:41:49 sessSendVerb: Error sending Verb, rc: -50
>>>03/24/99 20:41:49 TcpFlush: Error 32 sending data on
Tcp/Ip
>>>socket 4.
>>>03/24/99 20:41:49 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:41:49 ANS1809E Session is lost; initializing session
>>>reopen
>>>procedure.
>>>03/24/99 20:42:04 ANS1810E ADSM session has been reestablished.
>>>[end of ErrorLog]
>>>
>
>
>Richard C. Dempsey email: dempsey AT kodak DOT com
>Public Online Services pager: 716-975-3539
>11th Floor, Bldg 83, RL phone: 716-477-3457
>Eastman Kodak Company
>Rochester, NY 14650-2203
>
>
Richard C. Dempsey email: dempsey AT kodak DOT com
Public Online Services pager: 716-975-3539
11th Floor, Bldg 83, RL phone: 716-477-3457
Eastman Kodak Company
Rochester, NY 14650-2203
|
